• Worked with product, risk, and data science teams to understand how customers interacted with PayPal Wallet, translating those insights into the features and data signals our recommendation and fraud models needed.

• Designed the backend service flow using Python, Flask, and AWS EC2, making sure the recommendation service could respond quickly enough to support real-time user personalization across multiple Wallet surfaces.

• Set up Kafka pipelines to stream transaction and behavioral events, processing over 1M events per day so our ML services always had fresh data to work with.

• Built PySpark feature workflows that captured patterns like transaction velocity, device consistency, and location anomalies, which led to 8–12% better accuracy in both fraud and recommendation models.

• Developed the real-time scoring services with PySpark, Scikit-learn, and Flask APIs, and connected them to PayPal’s internal Risk and Identity systems using gRPC so they could make decisions based on the most current user context.

• Improved performance of the data layer by indexing key tables in PostgreSQL and caching high-use attributes in Redis, cutting service latency from 350 ms to around 120 ms during peak load.

• Containerized the services with Docker and deployed them through Kubernetes, which helped the platform scale smoothly during events where transaction volume spiked significantly.

• Built Jenkins pipelines to automate testing and deployments, adding unit tests, schema checks, and load testing so new model versions could be deployed quickly without breaking downstream systems.

• Created monitoring dashboards using Python, Tableau, and internal APIs, giving teams real-time visibility into model behavior, CTR trends, latency shifts, and early signs of model drift during A/B tests.

• Led end-to-end Oracle FLEXCUBE upgrade initiatives for 12+ banking clients, coordinating with client IT teams to plan schema changes, validate module compatibility, and execute production cutovers with zero disruption to daily banking operations.

• Designed and integrated 10+ Java-based REST APIs to support customer onboarding, account creation, and transaction approvals, reducing manual branch processing and improving operational consistency.

• Developed automated data migration scripts using Java, Shell scripting, and PL/SQL to securely transfer millions of customer and transaction records across environments, significantly improving migration speed and accuracy.

• Optimized system performance by tuning complex SQL queries and PL/SQL procedures, cutting critical query execution times by nearly 50% and ensuring timely completion of daily and end-of-day batch processes.

• Diagnosed and resolved 50+ production issues by analyzing SQL traces, FLEXCUBE application logs, and JVM metrics, addressing root causes that impacted transaction processing and system stability.

• Improved Java service performance by optimizing memory utilization, connection pooling, and thread management, enabling the platform to handle peak transaction volumes with fewer latency issues.

• Built Java- and SQL-based validation frameworks to verify Payments, Loans, and Limits modules after upgrades, reducing manual QA effort and catching defects earlier in the release cycle.

• Implemented performance monitoring routines using Oracle SQL Developer and diagnostic tools to track API response times and database activity, helping proactively identify performance risks before deployment.

• Worked closely with DBAs, QA teams, and client stakeholders to gather functional requirements, implement regulatory updates, and ensure FLEXCUBE configurations aligned with each bank’s operational workflows.